IA: WHITE COLLARS RUST TOO
A new index out of Tufts University has put numbers on a fear that has been circulating through newsrooms, product teams, and analytics departments for a while now: artificial intelligence is not coming for the routine tasks everyone expected. It is coming for the thinking.
For years, the comfort offered to office workers was simple: robots replace muscle, not mind. Automation ate assembly lines, not newsrooms. That comfort just collapsed. The American AI Jobs Risk Index, released in March by Digital Planet, the research center at Tufts' Fletcher School, ranked 784 U.S. occupations by their actual vulnerability to AI, not merely their theoretical exposure to it. The result is uncomfortable for anyone who makes a living writing, coding, or designing interfaces.
Writers and authors top the list, facing a 57.4% risk
of job loss within the next two to five years. Computer programmers follow
closely at 55.2%. Web and digital interface designers reach 54.6%, and editors
54.4%. Further down, but just as exposed, are operations research analysts,
space scientists, sociologists, technical writers, statisticians, political
scientists, database administrators, and public relations specialists, all
above 37%. Taken together, these are some of the best-paid, most prestigious
professions in the knowledge economy.
The Tufts team, led by dean Bhaskar Chakravorti, didn't settle for measuring how closely a task resembles something a language model can do. They combined fifteen years of labor market data with the most recent research on AI adoption — including the Anthropic Economic Index — to estimate not just what could be automated, but what is actually likely to be. The distinction matters: a task being technically reproducible by a model doesn't mean a company will choose to replace the person doing it. The index tries to capture that economic decision, not just the technical capability.
The aggregate picture is just as unsettling as the occupation-by-occupation detail. Under the median scenario, 9.3 million U.S. jobs are at risk, in a range running from 2.7 to 19.5 million depending on how fast the technology is adopted. Translated into wages, that's roughly $757 billion a year on the line. The researchers compare it, without much metaphor, to wiping an economy the size of Belgium off the map.
But the finding that should worry Silicon Valley most is geographic. The regions investing the most in building artificial intelligence — the San Francisco Bay Area, Boston, Washington D.C., Seattle — are also the ones facing the highest projected risk of workforce displacement from that very technology. Information, Finance and Insurance, and Professional, Scientific, and Technical Services concentrate the highest vulnerability by industry, well above the national average of roughly 6%. The people building the tool are standing in its line of fire too.
There is, however, one finding that breaks the expected narrative: the safest jobs right now are not the most sophisticated ones — they're the lowest paid. Roofers, dishwashers, and care aides face a displacement risk below 1%. Roughly 38% of American workers fall into that very-low-risk band, not because their work is inherently hard to automate, but because it's physical, tied to a specific place, or requires the kind of unpredictable judgment that current AI still handles poorly. The irony is hard to miss: economic protection no longer runs parallel to professional prestige. Sometimes it runs the other way.
Tufts' researchers also flag something subtler than a ranking: 33 "tipping point" occupations that currently sit at relatively low risk, under 10%, but could jump above 40% if AI adoption accelerates. Together, those occupations cover 4.9 million workers who feel safe today and, according to the model, might not be for much longer. It isn't a list of doomed professions; it's a list of professions riding on decisions companies haven't made yet.
These figures deserve the skepticism owed to any projection: they are adoption scenarios, not layoffs that have already happened, and the authors themselves note that future versions of the index will incorporate job-creation data alongside loss estimates. The history of technology is full of moral panics that never materialized, and of real transformations nobody saw coming in time. This one might turn out to be both.
What is clear is that the reassuring line that AI "frees humans up for higher-value work" starts to ring hollow when the list of the most exposed is made up, precisely, of the people doing the highest-value cognitive work that exists today: writing with precision, coding with rigor, designing with intent, analyzing with judgment. Chakravorti puts it in a line worth taking seriously: AI is no longer just automating the routine, it's moving up the ladder toward the cognitive and analytical work that defines high-skill, high-wage careers.
For anyone who makes a living writing, coding, or designing, the Tufts index doesn't announce the end. But it does shut the door on complacency. The question is no longer whether AI can do parts of your job. According to the data, in more than half of the cases, it already can. The question is what you're going to do about that before someone else decides for you.
 |
| Source: Fast Company Fall 2026 |
AND
THE JOBS AI HASN'T FULLY INVENTED YET
If the Tufts index described who might lose their job, there's another body of data, less cited but just as solid, that describes something different: the jobs artificial intelligence is already creating, and the ones it's likely to create in the next few years. The headline number, from the World Economic Forum's Future of Jobs Report 2025, is this: by 2030, the world will lose 92 million jobs, but gain 170 million. The net gain is 78 million jobs. This isn't PR optimism; it's the aggregate projection from more than a thousand employers surveyed worldwide.
The numbers are worth reading carefully, because the story isn't "AI creates more than it destroys" in the abstract — it's something more specific: it destroys tasks and creates roles. Within the AI and data-processing category itself, the report estimates 11 million new jobs against 9 million displaced, a modest but real net gain of 2 million in that segment alone. The rest of the positive balance comes from sectors AI transforms without directly replacing: renewable energy, autonomous vehicles, healthcare, education.
The list of the ten fastest-growing jobs by percentage through 2030 is telling. Big data specialists lead, with projected growth around 110%. FinTech engineers follow at 86%, and AI and machine learning specialists at 81%. Software and application developers and security management specialists round out the upper tier. Further down sit data warehousing specialists, autonomous and electric vehicle specialists, UI/UX designers, and Internet of Things specialists. None of these ten occupations, as they're defined today, existed even a decade ago.
But the more interesting data point isn't on that list — it's in how the roles that still don't have a stable name in HR manuals are forming right now. Three years ago, titles like "prompt engineer," "MLOps specialist," "AI agent orchestrator," or "AI product manager" barely existed as categories for mass hiring. Today they show up in job postings from banks, hospitals, government agencies, and startups alike, not just Big Tech. Roughly 40% of employers surveyed by the World Economic Forum say they plan to create new roles specifically tied to implementing and overseeing AI systems, and two-thirds expect to hire AI-skilled talent in the coming years.
Some labor-market analysts have started calling this phenomenon the "orchestration economy": instead of framing it as human versus machine, the competitive edge shifts to people who can direct AI systems with judgment, verify their output, and translate it into business decisions. It isn't a job that replaces the model; it's a job that exists precisely because the model, left unsupervised, makes costly mistakes. Demand for these "orchestrators" is growing faster than for almost any traditional technical profile.
There's also a less-cited but equally important pattern: AI doesn't just generate technical roles, it also reinforces demand for work that depends on human judgment, empathy, or physical presence — precisely the capabilities today's models handle worst. Nursing, higher education teaching, environmental engineering, and frontline occupations like delivery, construction, and farming are among the roles projected to add the most jobs in absolute terms through 2030, according to the same report. The paradox is elegant: the more AI automates routine cognitive work, the more valuable the occupations become that depend on what the machine still can't replicate.
The wage data backs up that reading. In the United States, the Bureau of Labor Statistics projects computer occupations will grow roughly 12% between 2024 and 2034, compared to just 3% for employment overall. Job postings that require AI skills already pay, on average, noticeably higher wages than equivalent postings that don't, and their numbers are growing at a rate that outpaces conventional openings by a wide margin. This isn't a market that rewards anyone who mentions "artificial intelligence" on a résumé; it rewards those who can demonstrate they know how to work with it, correct it, and stand behind its results.
These projections deserve the same skepticism as any other. The Tufts researchers behind the risk index we covered a few weeks ago have themselves announced that future versions of their model will incorporate job-creation data too, precisely because analysis of the upside is still thinner and less mature than analysis of the risk. The World Economic Forum's optimism doesn't cancel out Tufts' unease; it complements it. Both studies describe the same transformation from different angles — one measures what's lost, the other what could be gained — and neither has the final word on how the outcome will actually be distributed.
What does seem consistent across every source is the direction of change: nearly two-fifths of today's job skills will be transformed before 2030, and the companies winning this transition aren't the ones laying off fastest — they're the ones redesigning roles fastest. The question this hopeful side of the debate raises isn't whether there will be work after AI. It's whether each person will manage to move in time toward the occupation that doesn't have a name yet, before someone else defines it first.
No hay comentarios.:
Publicar un comentario